how can I reach the exact same results as pvar2 code with using GMM

[Ibrahim Kesici]

Hi all,

I estimate panel VAR using the pvar2 command which uses the GMM estimator, but I do not get the same results as the GMM.

For example,

pvar2 x y, fod

and

gmm (h_x - {xb:L.h_x L.h_y}), inst(L1.x L1.y, nocons) winit(i) wmatrix(robust) vce(unadjusted)

give exactly same for results of the first equation of panel VAR. (h_x and h_y are Helmert transformed version of x and y for option "fod")

But if I add instlags(4 5) to pvar2 model [ pvar2 x y, fod instlags(4/5) ] for instance, I cannot get the same results with GMM.

gmm (h_x - {xb:L.h_xL.h_y}), inst(L(4/5).(x y), nocons) winit(i) wmatrix(robust) vce(unadjusted)

This, for example, doesn't give the same results. I've tried thousands of combinations already but it seems like I'm missing a point.

I would like some help with this issue. Thanks in advance.